Swarming agents for distributed pattern detection and classification
Swarming agents in networks of preferably physically distributed processing nodes are used for data acquisition, data fusion, and control applications. An architecture for active surveillance systems is presented in which simple mobile agents collectively process real-time data from heterogeneous sources at or near the origin of the data. System requirements are specifically matched to the needs of a surveillance system for the early detection of large-scale bioterrorist attacks on a civilian population, but the same architecture is applicable to a wide range of other domains. The pattern detection and classification processes executed by the proposed system emerge from the coordinated activities of agents of two populations in a shared computational environment. Detector agents draw each other's attention to significant spatio-temporal patterns in the observed data stream. Classifier agents rank the detected patterns according to their respective criterion. The resulting system-level behavior is adaptive, robust, scalable, and applicable to a wide variety of other situations, including surveillance, financial transactions, network diagnosis, power grid monitoring, and others.
1. A swarming agent architecture for the distributed detection and classification of spatio-temporal patterns representative of biological populations, comprising:
a population of mobile software agents in a computer network of processing nodes; and
a service software agent at each node, enabling the mobile agents to perform the following functions:
a) share information on an indirect basis,
b) coordinate activities across the entire biological population, and
c) detect and classify one or more of the spatio-temporal patterns based upon the information sharing, coordinated activities, or both; and
a processor to perform data acquisition data fusion, or control functions in accordance with the patterns to simulate the behavior of a biological population, and output the results of the simulation to predict what might happen if the patterns actually arose.
2. The architecture of claim 1 , wherein the spatio-temporal patterns occur in a heterogeneous real-time data stream.
3. The architecture of claim 1 , wherein one or more of the spatio-temporal patterns are associated with a biological population.
4. The architecture of claim 1 , wherein one or more of the spatio-temporal patterns are associated with an environment.
5. The architecture of claim 1 , wherein one or more of the spatio-temporal patterns are associated with an event.
6. The architecture of claim 1 , wherein one or more of the spatio-temporal patterns are associated with a network.
7. The architecture of claim 6 , wherein the network is related to communications or energy.
8. The architecture of claim 1 , wherein one or more of the spatio-temporal patterns are associated with surveillance.
9. The architecture of claim 1 , wherein the definition of a pattern depends on the particular application of the detection mechanism.
10. The architecture of claim 1 , including detector agents that search for spatio-temporal structures and classifier agents that classify the detected patterns according to a predetermined classification scheme.
11. The architecture of claim 10 , wherein the detector agents are attracted to locations which other detectors identified as part of a potential pattern, and repelled by large concentrations of other detector agents.
12. The architecture of claim 10 , wherein the pattern classification scheme expresses a particular spatial or temporal characteristic.
13. The architecture of claim 1 , including places to represent the structure of the problem space in which the agents coordinate their activities.
14. The architecture of claim 13 , wherein the places represent a physical space.
15. The architecture of claim 13 , wherein the places represent a temporal space.
16. The architecture of claim 13 , wherein the places represent an abstract space.
17. The architecture of claim 13 , further including a place agent at the location of the processing to gather data from local sources and for storage to provide ready access.
18. The architecture of claim 13 , wherein the place agent provides application agents with a unified interface to the data streams covered by the place.
19. The architecture of claim 13 , wherein, the agents are operative to change focus by moving to a neighboring place, or by incrementing or decrementing an internal variable.
20. The architecture of claim 1 , further including a data provider giving mobile agents access to data characteristics of the segment of the domain represented by the processing node.
21. A computer system programmed to implement the architecture of claim 1 .
22. A swarming agent architecture for the distributed detection and classification of spatio-temporal patterns representative of biological populations, comprising:
a population of mobile software agents in a computer network of processing nodes, including:
detector software agents that search for the spatio-temporal patterns, and
classifier software agents that classify the detected patterns according to a predetermined classification scheme; and
a processor to perform data acquisition data fusion, or control functions in accordance with the patterns to simulate the behavior of a biological population, and output the results of the simulation to predict what might happen if the patterns actually arose.
23. The architecture of claim 22 , wherein the spatio-temporal patterns occur in a heterogeneous real-time data stream.
24. The architecture of claim 22 , wherein new classification schemes are introduced at different times through new agent populations.
25. A computer system programmed to implement the architecture of claim 24 .
26. The architecture of claim 22 , wherein one or more of the spatio-temporal patterns are associated with an environment.
27. The architecture of claim 22 , wherein one or more of the spatio-temporal patterns are associated with an event.
28. The architecture of claim 22 , wherein one or more of the spatio-temporal patterns are associated with a network.
29. The architecture of claim 28 , wherein the network is related to communications or energy.
30. The architecture of claim 22 , wherein one or more of the spatio-temporal patterns are associated with surveillance.
31. The architecture of claim 30 , wherein the pattern classification scheme expresses a particular spatial or temporal characteristic.
32. The architecture of claim 22 , wherein the definition of a pattern depends on the particular application of the detection mechanism.
33. The architecture of claim 22 , wherein the detector agents are attracted to locations which other detectors identified as part of a potential pattern, and repelled by large concentrations of other detector agents.
34. The architecture of claim 22 , including places to represent the structure of the problem space in which the agents coordinate their activities.
35. The architecture of claim 34 , wherein the places represent a physical space.
36. The architecture of claim 34 , wherein the places represent a temporal space.
37. The architecture of claim 34 , wherein the places represent an abstract space.
38. The architecture of claim 34 , further including a place agent at the location of the processing to gather data from local sources and for storage to provide ready access.
39. The architecture of claim 34 , wherein the place agent provides application agents with a unified interface to the data streams covered by the place.
40. The architecture of claim 34 , wherein, the agents are operative to change focus by moving to a neighboring place, or by incrementing or decrementing an internal variable.